Nebius Third-Quarter Revenue Jumps 571% to $981 Million as $27 Billion Capex Spooks Investors

Nebius Group N.V. reported a 571% surge in third-quarter revenue, reaching $980.8 million as demand for artificial intelligence cloud infrastructure continues to accelerate globally. The explosive top-line growth highlights the immense scale of corporate investment in compute capacity across the technology sector.[3]

Despite the historic revenue jump, the company's stock plunged immediately following the earnings release. Shares fell between 15% and 20% in active trading as markets digested the broader financial outlook and the implications of the revised guidance for the upcoming fiscal year.[2][5][7]

The severe market reaction stems from a combination of soaring capital expenditures and a downward revision to the company's near-term sales trajectory. Investors focused heavily on the massive costs required to sustain the current growth rate rather than celebrating the immediate top-line beat.[1]

Nebius raised its 2026 capital expenditure forecast to a staggering $27 billion. The sheer scale of the spending, including a $7.3 billion outlay that immediately sparked concern, overshadowed the quarterly revenue beat and alarmed institutional shareholders looking for a path to sustained profitability.[2][5]

Furthermore, the company cited persistent colocation delays as a primary operational headwind. These physical bottlenecks forced Nebius to push its 2026 revenue forecast to the low end of its previous guidance, signaling that growth may decelerate in the coming quarters despite strong underlying demand.[1][2]

The Mechanics of AI Infrastructure

To understand the market's hesitation, it is necessary to examine the physical realities of building artificial intelligence cloud infrastructure. The sector operates on a fundamentally different capital cycle than traditional enterprise software, requiring massive physical footprints and unprecedented energy resources.

Companies like Nebius do not merely sell digital services; they lease access to massive clusters of high-performance processors. These specialized graphics processing units are strictly required for training and running complex artificial intelligence models at a commercial scale, making them the foundational currency of the industry.

Acquiring these advanced processors requires immense upfront capital, which is directly reflected in the $27 billion expenditure target for the 2026 fiscal year. Securing the hardware from manufacturers is the most expensive phase of the operation, requiring billions in committed capital before a single server is deployed.[2]

However, purchasing the silicon is only the first step in the deployment pipeline. The hardware generates no revenue while sitting in a warehouse awaiting installation, making deployment speed a critical metric for financial success in the highly competitive cloud infrastructure market.

The servers must be housed in specialized data centers equipped with advanced liquid cooling and massive power capacities. This integration process, known as colocation, is highly complex and geographically constrained by available energy resources and local zoning regulations.

When colocation facilities face construction slowdowns or power grid connection delays, the processors cannot be brought online. This directly delays the revenue they would otherwise generate for the cloud provider, trapping billions of dollars of capital in unproductive assets while depreciation clocks begin ticking.[1][2]

The Colocation Bottleneck

The specific colocation delays highlighted by Nebius represent a broader bottleneck in the global artificial intelligence supply chain. Physical infrastructure is struggling to keep pace with the unprecedented digital demand from enterprise customers rushing to deploy their own models.[1]

While silicon manufacturers have successfully scaled up production to meet the surge in orders, the facilities required to host the chips remain scarce. Building a modern data center is a multi-year endeavor fraught with regulatory hurdles, supply chain constraints, and intense competition for prime real estate.

These facilities require extensive permitting, specialized construction techniques, and massive power allocations from regional grids. This creates a severe timeline mismatch with the rapid delivery schedules of hardware vendors, leading to inevitable deployment backlogs across the entire cloud computing industry.

For Nebius, this means that while customer demand remains exceptionally high, the company cannot recognize the associated revenue until the servers are fully operational. The compute capacity remains stranded in the interim, frustrating growth projections and delaying the return on the company's massive hardware investments.[1][2]

Consequently, the delayed deployment pushes expected 2026 sales to the lower bound of the company's earlier projections. This shift in timing was enough to spook short-term investors looking for immediate returns on the massive capital outlays, triggering the sharp sell-off in the company's shares.[1][4]

Prefunding the Build

The financial structure of these infrastructure agreements adds another layer of complexity to the company's earnings profile. The cash flow dynamics differ significantly from traditional enterprise cloud contracts, shifting risk between the provider and the client in unique ways.

According to analysis from Morningstar, Nebius operates on a model where customers effectively prefund the build-out of the infrastructure. Clients commit capital upfront to secure future access to scarce compute clusters, guaranteeing their place in line as capacity slowly comes online.[6]

This prefunding model provides Nebius with the immediate liquidity needed to purchase billions of dollars in hardware. It ensures that the massive capital expenditures are backed by guaranteed future demand, mitigating some of the financial risk associated with building speculative data center capacity.[6]

However, this creates a scenario where the bill for the physical infrastructure comes due later, heavily weighting the company's balance sheet toward future delivery obligations. The execution risk remains entirely on the provider to bring the servers online and fulfill the contracts they have already signed.[6]

Investors are currently weighing the certainty of this prefunded revenue against the operational risks associated with deploying the hardware at scale. Any delay in deployment strains the underlying financial model and delays the recognition of top-line growth, creating volatility in the company's quarterly earnings reports.[4][6]

The Capex Conundrum

The sheer scale of the $27 billion capital expenditure forecast underscores the capital-intensive nature of the artificial intelligence cloud race. Providers are locked in an expensive battle for market share that requires unprecedented levels of investment just to maintain a competitive baseline.[2]

Spending $7.3 billion in a single quarter highlights the aggressive pace at which Nebius is attempting to secure its position. The company is prioritizing long-term capacity over short-term margin preservation, a strategy that tests investor patience during periods of macroeconomic uncertainty.[5]

Market analysts often view such massive outlays with caution, as they require sustained, multi-year demand to generate an acceptable return on investment. The depreciation schedules on this specialized hardware are unforgiving if utilization rates drop or if next-generation chips render current models obsolete.[4][5]

If the broader artificial intelligence market were to cool, companies carrying tens of billions in infrastructure debt could face severe financial strain. The hardware cannot easily be repurposed for traditional computing tasks, creating a significant overhang if the anticipated enterprise demand fails to materialize.

AI

Yet, scaling back investment is equally risky, as it would cede critical ground to competitors in a market defined entirely by available compute capacity. Providers must spend heavily simply to remain relevant in the infrastructure arms race, leaving them with few alternatives to the current spending trajectory.

Looking Ahead

The stark divergence between Nebius's 571% revenue growth and its 20% stock decline illustrates the tension between operational success and market expectations. Wall Street is demanding flawless execution alongside hyper-growth, punishing any sign of physical friction in the deployment pipeline.[3][7]

The cited financial reports and preliminary earnings coverage did not include direct quotations from Nebius executives regarding the exact timeline for resolving the colocation delays. The market is operating largely on the revised guidance figures and capital expenditure targets provided in the initial release.[1][2][3][4][5][6][7]

For the remainder of the 2026 fiscal year, the company's valuation will likely hinge on its ability to clear these physical bottlenecks. Execution on the ground is now more critical than securing new orders from enterprise clients, as the backlog of prefunded demand continues to grow.[1][6]

If Nebius can accelerate the deployment of its hardware, the delayed revenue will eventually flow through to the income statement. The prefunded demand guarantees the sales, provided the infrastructure is built and connected to the grid in a timely manner.[6]

Until those physical constraints are lifted, however, the massive capital expenditures will continue to weigh heavily on investor sentiment. The next verifiable checkpoint will be the fourth-quarter deployment figures, which will reveal whether the delays are temporary or structural challenges for the broader industry.[1][2]